Shipley, T.H., Ogawa, Y., Blum, P., and Bahr, J.M. (Eds.), 1997Proceedings of the Ocean Drilling Program, Scientific Results, Vol. 156

23. SYNTHESIS OF THE BARBADOS DÉCOLLEMENT SEISMIC REFLECTION RESPONSEFROM DRILLING-BASED GEOPHYSICAL OBSERVATIONS AND PHYSICAL PROPERTIES 1

Thomas H. Shipley,2 Gregory F. Moore,3 Harold J. Tobin,4 and J. Casey Moore5

was to8 isick faultnhole. A

neratinsence of polarityd reve den-eismict to h

ling don very lowich areessureveformt-trendingollement,

33

ABSTRACT

A three-dimensional seismic reflection experiment has revealed a distinctive distribution of waveforms and amplitudesassociated with the décollement reflection in the Barbados Trench. One objective of Ocean Drilling Program Leg 156calibrate this seismic reflection response, which was thought to be related to variations in fault zone properties. Site 94 char-acterized by a normal polarity reflection that is correlated to the décollement. Sampling and logging reveal a 40-m-thzone, with density that increases across the zone as the clay mineralogy changes from smectite rich to illite rich dowsynthetic seismogram, derived from the logging data, shows that the lithology-driven density shift is responsible for gegthe normal polarity reflection even though the log, packer, and long-term pressure measurements confirm the preabnormally high pressures. The objective of Site 947, to characterize the more common, high-amplitude negativewaveform, was not completed due to technical difficulties. However, a density log, run at the beginning of the cruise, diala 14-m-thick 1.6 g/cm3 layer 540−554 m below seafloor (mbsf). The bit reached within 60 m of the décollement, and thesity log went to 565.5 mbsf. The low-density layer correlates with a fault splay off the décollement, having similar scharacteristics to that of the negative polarity décollement and can serve as an analogue. Although it is unsatisfying noavedirectly calibrated the décollement at Site 947, the fault splay revealed characteristics that are consistent with modeebefore the cruise that required a 12- to 14-m-thick low impedance layer to account for the observed waveform. Thesedensities are probably created by introduction of additional fluids and fluidized sediments during hydrofracturing, whmaintained by lithostatic pressure. The requirement for extremely low density indicates that a >12-m-thick lithostatic prdhydrofractured zone occurs within the thicker décollement. The variations in the amplitude of the negative polarity warelate mainly to the amount of enhanced fluid content. The mapped pattern of amplitudes suggests a broad northeaschannel that possibly focused fluid transport along the décollement. However, numerous fault splays intersect the décwhich may cause barriers that prevent rapid fluid flow along the décollement.

ariea

93;lude

rt re-i andf lowcog-are dé-e ofell-

f thes ofity

oth-de,rel-d tone

ningsso-94;

3-Dt var-cts ofro- tot.-D

r of the

INTRODUCTION

In the last decade, observations of fluid seeps in subduction zones,associated with heat and mineral transfer during pore-water advectionand fluid-rock chemical exchange, place renewed emphasis on theimportance of fluids in accretionary prisms (Kulm et al., 1986; Car-son et al., 1990; Kastner et al., 1991; Moore and Vrolijk, 1992). Theeffects of fluids on fault strength are particularly important in conver-gent margin toe-of-slope regions where forced tectonic dewateringproduce exceptional rates of porosity reduction and fluid expulsion.There is mounting evidence from accretionary prisms at convergentmargins that fluids reach pressures far higher than hydrostatic, caus-ing localized weakening, faulting, and even hydrofracturing (J.C.Moore et al., 1995b; Fisher et al., 1996). Multiple cross-cutting setsof veins in accretionary prism rocks indicate crack generation andmineralization induced by cyclic lithostatic fluid pressures and fluidflux (Sibson, 1981). Seismic reflection profiles often provide a coher-ent mappable reflection associated with the décollement as imbeneath the Costa Rica, Nankai and Barbados accretionary pand on fault ramps off Oregon that are closely associated with thcollement (e.g., Shipley and Moore, 1986; Shipley et al., 1990 B

1Shipley, T.H., Ogawa, Y., Blum, P., and Bahr, J.M. (Eds.), 1997. Proc. ODP, Sci.Results, 156: College Station, TX (Ocean Drilling Program).

2Institute for Geophysics, University of Texas, Austin, TX 78759-8397, U.S.A.tom@utig.ig.utexas.edu

3Department of Geology and Geophysics, University of Hawaii, Honolulu, HI96822-2285, U.S.A.

4Department of Geophysics, Stanford University, Stanford, CA 94305-2215, U.S.A.5Earth Sciences, University of California, Santa Cruz, Santa Cruz, CA 95064-0001,

U.S.A.

UHYLRXV&KDSWHUUHYLRXV&KDSWHU 7DEOHRI&7DEOHRI&

gedsms, dé-ngs

and Westbrook, 1991; Moore et al., 1990; Moore and Shipley, 19Tobin et al., 1993; J.C. Moore et al., 1995a). These studies concthat the presence of unusual volumes of fluids are in a large pasponsible for the seismic response of fault zones. In the NankaBarbados cases, the seismic waveforms indicate the presence odensity associated with the décollement, and in both areas, renized changes in the waveform along individual seismic lines caused by spatial variations in thickness and porosity within thecollement. In this paper we explore the specific seismic responsthe décollement off Barbados. Here there is both exceptional wcontrol and three-dimensional (3-D) seismic data.

The Barbados Trench is the eastward extent of subduction oNorth America Plate beneath the Caribbean. The primary locushear is revealed in particular detail by the reflectivity and continuof the fault-plane seismic reflection, enhanced by a setting whereer nearby reflecting horizons are often absent or of low amplituthus minimizing interference (Westbrook and Smith, 1983). This atively simple setting and seismic response of the décollement lea 3-D seismic reflection study that yielded a map view of fault-plareflection peak amplitude. This map spurred speculation concerporosity, fluid pressure, fluid flow, and even strength and stress aciated with the plate-boundary thrust (Figs. 1, 2; Shipley et al., 19G. F. Moore et al., 1995; Bangs et al., 1996). The results of thesurvey demonstrate that the seismic response of the décollemenies at mappable scales, leading to the question of just what aspethe fault properties produce the seismic signal. Ocean Drilling Pgram (ODP) Leg 156 (June–July 1994) provided an opportunitycalibrate the seismic waveform to the geology of the décollemen

Even the small area of the décollement illuminated by the 3seismic survey (about 100 km2 of the 125-km2 survey area) reveals ascale of heterogeneity that requires calibration to a finite numbedrill holes and in situ measurements (Fig. 1). Understanding how

293RQWHQWVRQWHQWV 1H[W&KDSWHU1H[W&KDSWHU

T.H. SHIPLEY, G.F. MOORE, H.J. TOBIN, J.C. MOORE

600

650

700

750

800

1000 1100 1200 1300 1400

0 1 2 -9.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 4.3

-58˚ 44' -58˚ 42'

15˚ 32'

541 542

671

675676

947948

949

Kilometers Seismic decollement amplitude

Common midpoint

Line

Figure 1. Map showing peak amplitude of the seismic expression of the décollement. Site 948 is located in one of the rare areas of positive polarity (orange).The basis for the map of peak amplitude is 103 seismic reflection lines, each 50 m apart. Lines were interpolated between each other, yielding a 25-m spacingbefore 3-D post-stack migration. These 205 lines with 15-m common reflection midpoints (CMPs) were then interpreted, and the amplitudes extracted. A 20-mmedian smoothing filter was applied to the map.

h d

r

gs

t

8

rtitp

ac

plyd tocon-d ofres).mi-DP

ve-elsn-edm al.,

t isick

ayerrdeap-e isthe

seismic reflection expression of the décollement relates to the pcal properties of the fault is fundamental to the use of the seismicfor regional studies. In many respects the seismic data reveal thetial variations of the décollement previously suspected but nevetually observed. Many deductions have been made concerningsignificance of the map, drawing on the general understandinhow young fault systems develop, on the specifics of the Barbadogion, and on the results of previous drilling, which showed thatdécollement is a 30- to 50-m-thick layer of clay-rich sediments wscaly fabric and stratal disruptions (J.C. Moore et al., 1982, 19These previous drilling data revealed a close association of thecollement with decreasing smectite downhole. We present hesummary of the latest drilling efforts that elucidate fault properand the origin of the seismic reflection response, which is imporfor quantitative use of the seismic signature as a regional maptool (Figs. 3, 4).

BARBADOS SEISMIC DATA

The primary improvements afforded by the 3-D data set off Bbados is two-fold: it more properly images the décollement, redu

294

ysi-ata

spa- ac- the of re-heith8). dé-e aesanting

r-ing

out-of-plane effects that could cause waveform modifications simbecause of improperly placed energy, and it provides a methomap the décollement (and fault physical properties) over a large tiguous area. The 3-D seismic volume, after processing, consiste205 lines, each 25 km long and 25 m apart (see captions to figuThe original data were collected in June 1992, and the first 3-D grations, completed a year later, were the basis for refining the OLeg 156 drilling program. Since the drilling, estimates of seismic locities have been improved using post-stack 3-D trial velocity panalong with the two vertical seismic profile (VSP) experiments coducted during Leg 156. The velocity improvements were combinwith a better migration algorithm to improve images of intraprisstructures (Moore et al., Chapter 20, this volume; A. Teagan etunpubl. data).

The most common waveform associated with the décollemena compound reflection created from a layer, usually 12–14 m thbased on general seismic modeling (Shipley et al., 1994). This lhas lower impedance (velocity × density) than the sections above obelow. The intervening negative lobe of the waveform’s amplituand position were digitized (latitude, longitude, and depth) for mping purposes. It was also discovered that the reflection amplituda first-order indication of the relative impedance magnitude of

SYNTHESIS OF DÉCOLLEMENT SEISMIC REFLECTION RESPONSE

.

600

650

700

750

800

1000 1100 1200 1300 1400

-5500

-5500

-5500

-5500

-540

0

-540

0

-5400

-5400

-5400

-5400

0 1 2 -5650 -5600 -5550 -5500 -5450 -5400 -5350 -5300

-58˚ 44' -58˚ 42'

15˚ 32'

541 542

671

675676

947948

949

Kilometers Seismic decollement structure with 10-m structural contours

Common midpoint

Line

Figure 2. Structural contours of the décollement at a 10-m contour interval. Whereas there are some relations to the amplitude map, no overall pattern emergesFor instance the high-amplitude northeast-trending “channel” does not occupy a structural high, nor do the positive polarity patches occupy lows. These con-tours were produced by digitizing the 204 seismic lines.

n

ilit m

icde fnot

ce of

lle- po-

ent,iteandm/sceanure,ppedl.,

hichedns,theally

t sur-zoneent

layer because thin-bed tuning modifies the observed amplitudes lessthan 25% for a thickness >5 m (Shipley et al., 1994; Bangs et al.,1996). We will refer to this waveform as the negative polarity reflec-tion. One-dimensional forward models indicate a maximum velocitydecrease of 200 m/s in a layer about 12 m thick (Fig. 5). More typi-cally, the velocity decrease is 80–100 m/s. These models accouboth variations in thickness and interference tuning and assumHamilton-type (Hamilton, 1978) velocity-density function. We wexpress the impedance in terms of velocity, which is a more intuunit, assuming the conventional velocity-density relationshipHamilton (1978). The most important observation is that the seisresponse is from a layer with an impedance drop across theboundary and a step up in impedance at the lower boundary.

Whereas the negative polarity waveform is most characteristthe décollement, there is a second well-defined waveform referreas the positive polarity reflection. It is produced by an interface whthere is a step up in impedance and a geologic model where theis an interface juxtaposing lower velocity-density rocks of the haing wall with higher velocity-density rocks in the foot-wall, with n(seismically) detectable fault layer. Given perfect conditions,

t fore aliveofic

top

of toreaultg-

he

seismic source used in this experiment should detect the presena layer as thin as about 5 m (Shipley et al., 1994).

A map produced by digitizing the peak amplitude of the décoment seismic reflection on the 205 lines reveals that the negativelarity reflection covers 96% of the mapped extent of the décollemeven limiting the area to the well-defined décollement west of S949. A portion of the map is produced in Figure 1 where a broad bof high negative polarity amplitudes (corresponding to the >100 velocity decrease) trend northeast, not parallel to the trench, oplate fabric, or associated with the décollement surface structsuch as closure. An early interpretation suggested the pattern mahigher porosity and probably higher permeability (Shipley et a1994). Could the impedance trend represent a “channel” along wfluid flow is enhanced, or does it just represent a static fluid-fillpart of the décollement? Regions of positive polarity reflectiowhere no layer is detected, comprise only a small portion of mapped décollement (Fig. 1). These were interpreted as normcompacted sediments that may be stronger patches on the faulface, having analogies to fault characteristics in the seismogenic (Pacheco et al., 1993). A structural map of the reflection décollem

295

T.H. SHIPLEY, G.F. MOORE, H.J. TOBIN, J.C. MOORE

5 kmSea Floor

Volcanic basement 6 km

Site 947 Site 948Site 949

Common midpoints944 141912281099

1 Km

988

Figure 3. Seismic cross section of three Leg 156 drilling sites. Section was constructed through the three drill sites as shown by the dashed line in Figure 1. Den-sity logs are illustrated for Sites 947 and 948. Dark blue boxes in décollement represent screened intervals for packer and long-term CORK experiments at Sites948 and 949 (Zwart et al., Chapter 24, this volume; Becker et al., Chapter 19, this volume; Foucher et al., Chapter 18, this volume). The seismic acquisition andprocessing steps are detailed elsewhere (Shipley et al., 1994; G.F. Moore et al., 1995). Boxes at the three sites are sections expanded in Figure 4. Orange is highpositive amplitude, white is zero amplitude, and gray-to-black is increasing negative amplitude.

r

oi d

t

di

llh

rm

u

thres

gic

ity.thisSiteg-he ford thee peraseannubecitiescity bythetant,ug-ity,n-indi-r its

leill-mb-lay-

the

shows the general west dip and no simple correlation to the amplitudepattern (Fig. 2). Below we review the properties of the décollemas explored by drilling and then examine the significance of thecollement’s various seismic responses.

STRUCTURAL AND SEDIMENTOLOGIC ASPECTS

In this area of the Barbados Trench, four sites were drilled intodécollement by the Deep Sea Drilling Project (DSDP) and ODPSites 948/671, 541, 675, 949, and at two places seaward of the fthrust (Sites 543, 672; J.C. Moore et al., 1982, 1988). Defining theand bottom of the décollement is difficult given the incomplete recery and disrupted nature of the samples. The best drilling-defthickness is at Site 948 (the same location as Site 671) where abination of multiple-cored holes recovered samples with stratal ruptions and scaly fabric that, along with strain-rotation minealignment, indicate that the décollement is 40 m thick (Maltman eChapter 22, this volume; Housen et al., 1996). At other sites, it istimated to be at least 18 m thick, although the décollement coulmuch thicker because there was difficulty in drilling and sampl(Sites 541, 675, and 949). In all cases, it is thicker than the estim12 m thickness of the interval of active dilation. Within the décoment, the deformation is unequally distributed between zones witdiscernible deformation and sections with moderate to intense sfabrics. Thus, the décollement contains locally intense deformaand intervening (mostly undeformed) sections that are centimetemeters thick. The deformation pattern may be a series of anastoing shear-bands at all scales (Labaume et al., Chapter 4, this vol

High clay mineral content is characteristic of the stratigraphic stion associated with the décollement (70%−90% at Site 948, Meyerand Fisher, Chapter 27, this volume). At Sites 948/671 and 541 is a distinct shift downsection in the clay minerals from smectite (decreasing from about 60% to 20%) to illite rich within the faultsection (Fig. 6). Underwood and Deng (Chapter 1, this volume) gest that the transition in clay type results in a shear-strength dis

296

entdé-

the atontal topv-

nedcom-is-

ralal., es- be

ngatede- no

heartions toos-

me).ec-

ereichdug-con-

tinuity that causes localization of the décollement near the litholoboundary.

PHYSICAL PROPERTIES

Seismic response is primarily controlled by density and velocOther factors are also important, but are probably secondary for data set (e.g., Fresnel zones, noise, rigidity, and attenuation). 948/671 is well characterized from a combination of sampling, loging, and vertical seismic profiling, all relevant at the scale of tseismic waveform (Fig. 6). There are two rich data sets of velocitySite 948 (Fig. 6). One is a series of measurements made on boarship on cut cubes. These measurements were made about oncmeter through the fault zone, characterizing a velocity decredown-core of about 75 m/s measured along the core axis (Brückmet al., Chapter 8, this volume). The effects of open cracks in cut-csamples and sediment decompression should make these velolower than expected for in situ conditions. The second set of velodata is a wireline sonic log, although the quality was compromisedhole size and sticky clay, which caused caliper problems within décollement. Within the 40-m-thick fault zone, the wireline log daindicate a 50 m/s increase within the lower part of the décollemewhereas the overall profile, above and below the décollement, sgests nearly constant velocity. The wireline measurement of velocalthough relatively low quality, is more representative of in situ coditions than the discrete-sample laboratory measurements and cates little velocity effect associated with the décollement zone ohanging or foot walls.

The wireline density log, which is very susceptible to poor hoconditions, yielded unreliable data. However, the logging-while-dring (LWD) density profile shows an increase in density of 0.1 g/c3

through the 40-m-thick décollement, which was confirmed with laoratory measurements (Fig. 6). The logs reveal 1- and 3-m-thick ers of extremely low density (1.5–1.65 g/cm3), which are interpretedas open hydrofractures maintained at lithostatic pressure within

SY

NT

HE

SIS

OF

DÉ

CO

LLEM

EN

T S

EIS

MIC

RE

FL

EC

TIO

N R

ES

PO

NS

E

297

100 m

Faults Stratigraphic boundaries

2F3A

949B/C

(C)

ed on the LWD density logs (as in Fig. 7) at Sites 947 and 948. 947 (A), the well has been shifted 175 m east and 20 m deeperata. Blue regions represent interpretation of fluid-rich sectionsreted mainly from the drilling at Site 948 (B). The extent of the 50 m deeper within the underthrust section (Fig. 3).

947A100m

100m

100m

100 m100 m 948A

100m

Fluid charged parts of fault zones Zone of stratal disruptions and scaly fabricStratigraphic boundary withdecreasing smectite in unit 3A

2F3A

(A) (B)

??? ?

Figure 4. Extreme close-up of Sites 947, 948, and 949 seismic data. The top three are uninterpreted panels with inset synthetic seismograms basTheir location is shown as boxes in Figure 3. The location and depth of Site 947 are shown correctly in the bottom panel. In the top panel of Siteto better match the seismic data (see text). These data illustrate the scale difficulty in interpreting fault properties directly from the seismic dbased on the seismic response, modeling, and density logs. The green regions are the extent of the structurally disrupted décollement, interpsedimentologic Unit 2/3 boundary as defined at Site 948, is correlated eastward to Site 949 (C). Westward, this boundary may occur about

T.H. SHIPLEY, G.F. MOORE, H.J. TOBIN, J.C. MOORE

Tra

vel T

ime

(S)

7.1

7.2

7.3

CMP 700

500 m~

12 m

Line 630

Amplitude

Data

Model0

-50

1800

17501680

180016001710

180016801740

1730

180016701800

16601760

0 Model

CMP 1099

Data

18001650

1760

180017001750

180016201720

1800172017401700

1800

1760

CMP 932

Figure 5. Close-up of a portion of seismic Line 630 (top), which illustrates the variation in décollement waveforms. Bottom is the input to a synthetic model ofthe section from Bangs et al. (1996). Middle is the difference between the model-produced amplitudes and those observed, which is one measure of the model-ing quality. These models illustrate the extreme decrease in velocity of 80–200 m/s needed to account for the waveforms. The purple regions in the map of Fig-ure 1 represent velocity drops of >100–150 m/s, although most of the mapped region represents drops of about 100 m/s (green). The map in Figure 1 extendsfrom common midpoint (CMP) 944 to 1419, whereas the modeled data extend from CMP 700 to 1099. These models assumed a 1800 m/s velocity for the sec-tion immediately above the décollement and generated the best single-layer model to mimic the amplitude and shape of the waveform. Models assume thatvelocity and density vary in a matter similar to normally compacting sediments. Figure is modified from Bangs et al. (1996).

is

vm ce

st

n

senaes

do

m

to

ouv

t

l

ao-e-

ated

enr 9,the

ntl-itede at

tytnt

ityot-atadataa-

iss athist

tlen-omoc-bsf.

décollement (J.C. Moore et al., 1995b). The resistivity data confthat the data are reliable. Packer and long-term pressure meaments verify the presence of super-hydrostatic pressures here (Zet al., Chapter 24, this volume; Becker et al., Chapter 19, this ume). These thin layers are well below the resolution of the seissource waveform used in the seismic reflection survey and docontribute to the recorded seismic response (Fig. 6). On a large sthe density increase is a result of the lithologic transition and decring water-layer-rich smectite down core. Brückmann et al. (Chap8, this volume) confirms that the porosity calculated from the denis nearly constant through the décollement after adjusting for bound water that is associated with changing clay mineralogy (Broand Ransom, 1996). Whereas the porosity doesn’t change muchdensity is the driving force for the seismic signal here. A syntheseismogram created by convolving a wavelet with an impedafunction derived from the density and the linear increasing veloccreated a response very similar to the waveform from the seismicvey (Fig. 6; Moore et al., Chapter 20, this volume). The presencthe lithologic change in clay mineralogy is the cause of the impedaincrease and the resulting positive polarity waveform (Moore et Chapter 20, this volume; Tobin and Moore, Chapter 9, this volum

Site 947 provides other relevant information regarding acouproperties of the décollement. Here density logging stopped aboum above the seismic expression of the décollement (Fig. 7). The sity log showed a complicated profile in the bottom 50 m of the h(Fig. 8). The density decreases from 0.3 to 0.4 g/cm3 over a 14-m-thick interval. The only way to explain densities of 1.50–1.65 g/c3

at 540 mbsf is by fluids and fluidized sediments within hydrofratures (J.C. Moore et al., 1995b). The thick, low-density section is eily detectable seismically (see Figs. 4A, 8C), and, thus, we probalogged a fault splay above the décollement (see Discussion secthis chapter). The log is the best analogue we have to the commoccurring negative polarity reflection.

Unique data from the Oregon frontal thrust area has been useestimate velocity changes associated with effective pressure to euate the cause of the fault reflectivity observed there (Tobin et1994). In a similar study of five samples from the Barbados Trenabove and below the décollement, the general result is one of modest velocity change as the samples were cycled from in sit tonear-lithostatic fluid pressures (Tobin and Moore, Chapter 9, this ume). The maximum observed change in velocity was of the ordeabout 100 m/s, with velocities at in situ conditions equivalent to wireline sonic results. The velocity change associated with the barea on the map in Figure 1 is predicted to be about 100–150 m/ser by modeling (Fig. 5; Bangs et al., 1996). However, Bangs et

298

rmure-

wartol-ic

notale,as-terityhewn, theticce

ityur- ofcel.,).

tict 78en-le

c-as-blyion,nly

d toval-al.,chnly

ol-r ofhelueow-al.

(1996) used a Hamilton-type velocity-density function that predicts0.2 g/cm3 density decrease associated with the velocity change. Tbin and Moore (Chapter 9, this volume) found that the density dcrease is much smaller; their 100 m/s velocity decrease is associwith a density decrease of only about 0.1 g/cm3. Thus, this new datarequire the results of Bangs et al. (1996) to be modified to an evgreater negative impedance contrast. Tobin and Moore (Chaptethis volume) conclude that to achieve and maintain low densities, décollement must be hydrofractured at lithostatic pressure.

DISCUSSION

Calibration of the differing seismic responses of the décollemewas an original objective of the Leg 156 project. Operational difficuties at Site 948 did not allow sufficient time to return to complete S947, so we did not obtain a direct calibration of the high-amplitunegative polarity reflection. We did a much better characterizationSite 948, the site of the characteristic normal polarity reflection.

Site 947 as a Non-décollement HydrofracturedReference Site

At the base of Site 947, the density log is probably representativeof a hydrofractured layer, and if we correlate this to the seismic re-flection data, it provides a method to calibrate the negative polaritywaveform. The logged density revealed a large excursion to as low as1.5 g/cm3 in a 14-m-thick zone at 540–554 mbsf with the data qualiconfirmed by the resistivity log run (Fig. 8). This layer is of sufficienthickness and impedance to produce a reflection, if it has sufficielateral extent. At the projected position of Site 947, the low densdoes not correlate with any recorded seismic reflection (Fig. 4A, btom). The computed location of the hole relative to the seismic dhas several potential uncertainties. One issue is that the seismic is located with differential global positioning system (GPS) navigtion, whereas the JOIDES Resolution is located with conventionaldithered GPS. Also, the location of the drill bit, relative to the ship,uncertain. Altogether, the uncertainties could result in as much a200–300 m error in relative position. Because we had no VSP at site, the velocity-depth function is not known extremely well, buprobably is not in error by more than 25 m, and thus, there is litlikelihood that we actually reached the décollement (A. Teagan, upubl. data). Our best estimate for the depth to the décollement is frthe unpublished velocity study of Teagan, which estimates the velity at 1836 m/s, traveltime at 701 ms, and, thus, the depth at 644 m

SYNTHESIS OF DÉCOLLEMENT SEISMIC REFLECTION RESPONSE

ity

E

-0.8 -0.4 -0.0 0.4 0.8

Nomalized amplitude

D

1.6 1.8 2.0

Density (g/cm3)

C

0.6 0.8 1.0

Resistivity (Ωm)

B

1.6 1.7 1.8

Velocity (km/s)

A450

500

550

Dep

th (

mbs

f)

0 20 40 60 80

Smectite (%)

Figure 6. Site 948 downhole profiles through the structurally defined décollement from 490 to 530 mbsf (shaded region). A. The relative percentage of clayminerals in the <1-µm size fraction (Underwood and Deng, Chapter 1, this volume). B. Wireline sonic logs resampled at a 0.5-m interval. Triangles are velocmeasurements on laboratory samples, both along core axis and transverse (Brückmann et al., Chapter 8, this volume). C. LWD resistivity data resampled at a0.5-m interval. D. Density LWD profile resampled at a 0.5-m interval. Squares are laboratory measured index properties. E. Synthetic seismogram created byusing the density log, linear velocity function, and source wavelet derived from the field data (Moore et al., Chapter 20, this volume).

omnt pdrostaeonlyfire

e? ff olo

ec-th-3).e-ntd toect-dé-stssi-

-richec-lo-y of

ot-e-(Z.beingnt

tedg, The

he

We postulate that Site 947 penetrated the acoustic anomaly 200 m ei-ther to the east or west of the illustrated drill site (see relocation inFig. 4A, top). These anomalies have the same waveform characteris-tics as the negative polarity reflection associated with the décment. The anomalies located about 100 m above the décollemight be fault splays or flats on fault splays, preserved sedimebedding, or even relict portions of the décollement. If it had beensible to complete the reprocessing of the data (with the improvelocity functions and migration algorithm) before the drilling pgram, we would have searched for a drill site away from these lower anomalies. We interpret the fault zone to be at lithospressures with enhanced fluid content responsible for the low dty. Because this fault has characteristics very similar to the décment seismic waveform, we believe that much of the décollemea 10- to 14-m-thick, 1.5–1.65 g/cm3 section that was almost certaincreated by hydrofracturing and enhanced fluid content. This conmodeling estimates of thickness and impedance of the negativlarity reflection.

Site 948 as a Low Fluid Content Reference Site

The décollement is well sampled at Site 948, but how represtive is the lithology, physical properties and fluids for elsewherethis site the fault-plane reflection signature is not of a layer but ointerface downward to higher impedance. Because so much ocore-log data is associated with Site 948, there is a significant pbility of bias in using this site as the general analogue for the geo

lle-ent

aryos- ve--hal-ticnsi-lle-t is

ms po-

nta-At anourssi-gic

setting. How pervasive is the lithologic pattern and decreasing smtite through the décollement? The position of Site 948 is over a noreast-trending 2- to 3-km-wide, 150-m-high basement horst (Fig. Much of the relief is attenuated by preferential sediment infilling bfore deposition of the smectite-rich clays. However, the décollemeappears to be about 50 m lower in the section at Site 948 compareareas located over basement lows (e.g., Site 947; Fig. 3). The refling horizons, within the underthrust section, terminate against the collement near Site 948. This is not surprising since low-angle thrudo not precisely follow lithologic boundaries. This leaves the posbility that the physical properties of the foot-wall vary within themapped area as the fault juxtaposes smectite-poor and smectitesections in some areas (Site 948), but may be fully within the smtite-poor section in other areas. However, the smectite-poor lithogies have never been sampled above the active décollement in anthe DSDP or ODP coring sites; thus, the lithologic section in the fowall probably does not vary substantially, or this situation is rare. Dtailed mapping of the stratigraphic relationships are in progress Zhao et al., unpubl. data). We believe that the foot-wall is likely to less variable than the décollement hanging wall where faults splayfrom the décollement modify the overlying section at many differescales.

Site 948 was situated to calibrate the seismic signal associawith the normal polarity décollement (Fig. 3). Using the density lothe generated synthetic seismogram matches the survey data.density increase of about 0.1 g/cm3, a result of the downhole decreasein smectite and the reduction of interlayer water, is consistent. T

299

T.H. SHIPLEY, G.F. MOORE, H.J. TOBIN, J.C. MOORE

olle-icalbsf

ary con-nsity

nd at re-rmalpter0.1–SPrsionmaly. the

omalyg in-l lev-asingphicin alleng,ven am- low

tionr

ucedity isana- totals in-er.

arlyritical thewereted

om-

en-

rac- re-k

4

1.4 1.6 1.8 2.0 2.2

Density (g/cm3)

Site 948

1.4 1.6 1.8 2.0 2.2

Density (g/cm3)

Decollement(?)

Site 9470

50

100

150

200

250

300

350

400

450

500

550

600

650

Dep

th (

mbs

f)

Figure 7. Density logs for Sites 947 and 948 from the LWD tools. The stip-pled areas highlight the regions of density lows in the accretionary prism justabove the fault splay (at Site 947) and the décollement (at Site 948). Tsections have been attributed to elevated fluid pressures. This trend mcharacteristic of the décollement hanging wall. Shaded portion of Site 9the structurally defined décollement.

300

observed meters-thick low-density spikes are not detectable with thesource wavelet used in the experiment (Moore et al., Chapter 20, thisvolume). Just to the west, the seismic signature begins to develop intothe more characteristic low-impedance-layer negative polarity reflec-tion (although it is complicated in Figure 4B by interference from anearby overlying sedimentary horizon). It appears that the boundarybetween the two types of waveforms is broadly transitional, but as in-terpreted in Figure 4B, some places may be controlled by a fault splayoff the décollement that isolates fluid-charged regions of the décment and fluid migration westward. At this location, the geochemevidence for fluid flow is at the top of the décollement, at 490 m(Kastner et al., Chapter 25, this volume).

How representative is Site 948 of the overlying accretionprism? The best data comes from the logging because of its moretinuous nature and presumably near in situ measurements. Deinversions at Site 947 above the fault splay at 490–530 mbsf aSite 948 above the décollement from 390 to 500 mbsf (stippledgions of Figure 7) are interpreted as regions of moderately abnofluid pressures (J.C. Moore et al., 1995b; Tobin and Moore, Cha9, this volume). The density logs at both sites are shifted lower 0.2 g/cm3 from the projected downhole trend. In addition, the Vdata at Site 949 also suggest this inversion, so the density inveat the base of the accretionary prism may be a widespread anoHowever, the overall effect is probably minor, as at Site 948. Heredepressed densities associated with the broad fluid-pressure anwithin the prism increase the positive impedance contrast, helpincrease the amplitude of the positive reflector to detectable signaels. Even so, the positive impedance associated with the decresmectite downhole should occur as a primarily inherited stratigrareflection in the area, because the smectite-rich section is found holes that have reached the décollement (Underwood and DChapter 1, this volume). Although the general fluid-pressure drilow densities might increase the positive amplitude, the negativeplitudes should be depressed. However, in detail at Site 947, thedensity portion of the fault splay is surrounded by ~5-m-thick secof higher densities (1.9 g/cm3; Fig. 8, stippled region). These highedensities within the fault zone may be caused by shearing-indtectonic overconsolidation, although in this case, the actual densstill depressed below the overall downhole trend (Fig. 7). As an logue, this suggests that as the décollement undergoes moreshear (to the west), it is more likely to have higher densities, thucreasing the apparent impedance drop into a hydrofractured lay

Site 949 as an Incipient Décollement Site

Site 949 is situated in a more structurally and stratigraphicallycomplicated region very close to the deformation front. This site waschosen as an alternate to Site 947 when the time remaining was in-sufficient for drilling and sampling of Site 947. Traced to the south-east, the lower of two negative polarity reflections correlates to thedécollement at Site 948, although here the upper reflection is cletraced as a thrust to the seafloor (see Fig. 3). VSP data were cfor determining the position of the reflecting horizons to depth inhole. The sedimentologic units are correlated from Site 948 and also identified in the fragmentary stratigraphic section construcfrom rare core samples at Site 949. Until continuous logging is cpleted, Site 949 does not contribute to the waveform calibration.

CONCLUSIONS

Near the décollement, we conclude that the natural lithologic dsity variations are about 0.1 g/cm3 with associated small velocitychanges insufficient to account for the décollement reflection chateristics, except for the positive polarity reflection at Site 948. Aduction in density of 0.3 g/cm3 that was observed in the 12-m-thic

heseay be8 is

SYNTHESIS OF DÉCOLLEMENT SEISMIC REFLECTION RESPONSE

C

-0.8 -0.4 -0.0 0.4 0.8

Nomalized amplitude

B

1.4 1.6 1.8 2.0 2.2

Density (g/cm3)

A

0.8 1.0 1.2

Resistivity (Ωm)

500

550

Dep

th (

mbs

f)

Figure 8. Portion of (A) LWD resistivity log and (B) density log at Site 947, both resampled at a 0.5-m interval. Stippled zones might be shearing-induced density increases in the fault zone surrounding the ~12-m-thick, 1.6 g/cm3 hydrofractured section. How-ever, these densities are depressed relative to nor-mally expected downhole increases from compaction, so shear-induced strain is not necessar-ily the only explanation for the density profile. C. A synthetic seismogram created using the density log (Moore et al., Chapter 20, this volume).

t

o

a

Sei

rct>

p

t

t

ents948 re-lle-

ptlle-e-canidrm

er-asC-by

o

lle-x.

urear-

fault above the décollement at Site 947 is the appropriate magnito generate the common negative polarity seismic waveform. This ample evidence for super-hydrostatic fluid pressures from packer observations at both Sites 948 and 949, as well as in the clation obviation retrofit kit (CORK) data from these two sites (Fishet al., 1996; Zwart et al., Chapter 24, this volume). To modify nmally consolidated sediment and create such low densities requfluid and fluidized sediment injection into hydrofractures at lithostic pressures. The hydrofractures will be maintained only as longthe lithostatic pressures remain to drive open the fractures. Theseid-enhanced areas may be simple hydrologic compartments in thecollement. Given the presence of the hydrofractured layers at 948, hydrologic interconnectivity of the fault probably occurs evthough at seismic scales, the positive and negative polarity portof the décollement are distinctly different.

The observations of the fault splay at Site 947 and modeling of3-D data volume both suggest that the seismically responsive poof the décollement is about 12 m thick. Given the widespread ocrence of this thickness, it seems likely that the décollement almoserywhere (e.g., the negative polarity waveform) contains a layer m thick at near-lithostatic pressure. The acoustic signal revealsthe hydrologically active portion of the décollement is much thinnthan the structurally defined décollement (at least at Site 948).

All of these results suggest that the map of the reflection amtudes shown in Figure 1 relate the magnitude of the negative reflecwith the fluid content of the layer. Thus, the purple and blue portioof the map almost certainly represent a layer of the fault of particulalow density and, thus, higher porosity and permeability. If this is case, the map directly suggests that the more efficient fluid-flow paalong the décollement are directed northeasterly. However, if sudisruptions from young fault splays intersect the fluidized layer,

udeeretheircu-err-irest- as flu- dé-iten

ons

thetionur- ev-12

thater

li-tionnsrlyhethsbtlehe

regions may be more disconnected or divided into more compartmthan it appears. This may be one of the explanations for the Site positive polarity patch where limited mapping suggests a possiblelationship to high-angle faults that intersect or splay from the décoment and isolate a compartment (Fig. 4B).

Significantly higher resolution seismic data is needed to attemto define the detailed stratigraphy and deformation near the décoment and a few more carefully placed drill sites will eventually be rquired to fully calibrate the seismic response so the waveforms be inverted directly for porosity, permeability, and perhaps flupressure. As a first step, further calibration of the seismic wavefoshould occur in late 1996 when the JOIDES Resolution returns to thearea for additional logging.

ACKNOWLEDGMENTS

This work was supported by JOIDES member countries for opating the Ocean Drilling Program. The seismic reflection survey wsupported by NSF Grants OCE-9119336 (UH), OCE-9116350 (USC), and OCE-9116172 (UT). Additional support was provided USSAC for post-cruise interpretation. We also acknowledge twanonymous reviewers.

REFERENCES

Bangs, N.L.B., and Westbrook, G.K., 1991. Seismic modeling of the décoment zone at the base of the Barbados Ridge accretionary compleJ.Geophys. Res., 96:3853–3866.

Bangs, N.L., Shipley, T.H., and Moore, G.F., 1996. Elevated fluid pressand fault zone dilation inferred from seismic models of the northern Bbados Ridge décollement. J.Geophys. Res., 101:627–642.

301

T.H. SHIPLEY, G.F. MOORE, H.J. TOBIN, J.C. MOORE

s

ltp

ac

e

-

Su

c

)

E

M

.w

d

tive-ary

ardionill-

cou-

andsmic

94.dé-

nts,nary

nd

,

.,re-

ntal

lca-igh

Brown, K.M., and Ransom, B., 1996. Porosity corrections for smectite-richsediments: impact on studies of compaction, fluid generation, and tec-tonic history. Geology, 24: 843–846.

Carson, B., Suess, E., and Strasser, J.C., 1990. Fluid flow and masdeterminations at vent sites on the Cascadia margin accretionary prisJ.Geophys. Res., 95:8891–8897.

Fisher, A.T., Zwart, G.T., and ODP Leg 156 Scientific Party (T.H. Shipco-chief scientist), 1996. Relation between permeability and effecstress along a plate-boundary fault, Barbados accretionary comGeology, 24:307−310.

Hamilton, E.L., 1978. Sound velocity-density relations in sea-floor sements and rocks. J. Acoust. Soc. Am., 63:366−377.

Housen, B.A., Tobin, H.J., Labaume, P., Leitch, E.C., Maltman, A.J., ODP Leg 156 Shipboard Science Party, 1996. Strain decoupling athe décollement of the Barbados accretionary prism. Geology, 24:127−130.

Kastner, M., Elderfield, J., and Martin, J.B., 1991. Fluids in convergent mgins: what do we know about their composition, origin, role in diagenand importance for oceanic chemical fluxes. Philos. Trans. R. Soc. Lon-don A, 335:243−259.

Kulm, L.D., et al., 1986. Oregon subduction zone: venting, fauna, andbonates. Science, 231:561−566.

Moore, G.F., and Shipley, T.H., 1993. Character of the décollement inLeg 131 area, Nankai Trough. In Hill, I.A., Taira, A., Firth, J.V., et al.,Proc. ODP, Sci. Results, 131: College Station, TX (Ocean Drilling Program), 73−82.

Moore, G.F., Shipley, T.H., Stoffa, P.L., Karig, D.E., Taira, A., Kuramoto,Tokuyama, H., and Suyehiro, K., 1990. Structure of the Nankai Troaccretionary zone from multichannel seismic reflection data. J. Geophys.Res., 95:8753−8765.

Moore, G.F., Zhao, Z., Shipley, T.H., Bangs, N., and Moore, J.C., 19Structural setting of the Leg 156 area, northern Barbados Ridge ationary prism. In Shipley, T.H., Ogawa, Y., Blum, P., et al. (Eds.), Proc.ODP, Init. Repts., 156: College Station, TX (Ocean Drilling Program13−27.

Moore, J.C., Biju-Duval, B., Bergen, J.A., Blackington, G., Claypool, G.Cowan, D.S., Duennebier, F., Guerra, R.T., Hemleben, C.H.J., HussD., Marlow, M.S., Natland, J.H., Pudsey, C.J., Renz, G.W., Tardy, Willis, M.E., Wilson, D., and Wright, A.A., 1982. Offscraping and undethrusting of sediment at the deformation front of the Barbados RidDeep Sea Drilling Project Leg 78A. Geol. Soc. Am. Bull., 93:1065−1077.

Moore, J.C., Mascle, A., Taylor, E., Andreieff, P., Alvarez, F., Barnes,Beck, C., Behrmann, J., Blanc, G., Brown, K., Clark, M., Dolan, JFisher, A., Gieskes, J., Hounslow, M., McLellan, P., Moran, K., OgaY., Sakai, T., Schoonmaker, J., Vrolijk, P., Wilkens, R.H., and WilliamC., 1988. Tectonics and hydrogeology of the northern Barbados Ri

302 3UHYLRXV&KDSWHU3UHYLRXV&KDSWHU 7DEOHRI&7DEOHRI&

fluxm.

eyivelex.

di-

ndross

ar-sis

car-

the

.,gh

95.cre-

,

.,ong,

.,r-ge:

R.,F.,a,s,ge:

results from Ocean Drilling Program Leg 110. Geol. Soc. Am. Bull.,100:1578−1593.

Moore, J.C., Moore, G.F., Cochrane, G.R., and Tobin, H.J., 1995a. Negapolarity seismic reflections along faults of the Oregon accretionprism: indicators of overpressuring. J. Geophys. Res., 100:12,895−12,906.

Moore, J.C., Shipley, T.H., Goldberg, D., Ogawa, Y., and Leg 156 ShipboScientific Party, 1995b. Abnormal fluid pressures and fault-zone dilatin the Barbados accretionary prism: evidence from logging while dring. Geology, 23:605−608.

Moore, J.C., and Vrolijk, P., 1992. Fluids in accretionary prisms. Rev. Geo-phys., 30:113−135.

Pacheco, J.F., Sykes, L.R., and Scholz, C.H., 1993. Nature of seismic pling along simple plate boundaries of the subduction type. J. Geophys.Res., 98:14,133−14,159.

Shipley, T.H., and Moore, G.F., 1986. Sediment accretion, subduction, dewatering at the base of the trench slope off Costa Rica: a seireflection view of the décollement. J. Geophys. Res., 91:2019−2028.

Shipley, T.H., Moore, G.F., Bangs, N.L., Moore, J.C., and Stoffa, P.L., 19Seismically inferred dilatancy distribution, northern Barbados Ridge collement: implications for fluid migration and fault strength. Geology,22:411−414.

Shipley, T.H., Stoffa, P.L., and Dean, D.F., 1990. Underthrust sedimefluid migration paths and mud volcanoes associated with the accretiowedge off Costa Rica: Middle America trench. J. Geophys. Res.,95:8743−8752.

Sibson, R.H., 1981. Fluid flow accompanying faulting: field evidence amodels. In Simpson, D.W., and Richards, P.G. (Eds.), Earthquake Predic-tion: An International Review. Geophys. Monogr., Maurice Ewing Ser.Am. Geophy. Union, 4:593−603.

Tobin, H.J., Moore, J.C., MacKay, M.E., Orange, D.L., and Kulm, L.D1993. Fluid flow along a strike-slip fault at the toe of the Oregon acctionary prism: implications for the geometry of frontal accretion. Geol.Soc. Am. Bull., 105:569−582.

Tobin, H.J., Moore, J.C., and Moore, G.F., 1994. Fluid pressure in the frothrust of the Oregon accretionary prism: experimental constraints. Geol-ogy, 22:979−982.

Westbrook, G.K., and Smith, M.J., 1983. Long décollements and mud vonoes: evidence from the Barbados Ridge Complex for the role of hpore-fluid pressure in the development of an accretionary complex. Geol-ogy, 11:279−283.

Date of initial receipt: 1 August 1996Date of acceptance: 7 January 1997Ms 156SR-034

RQWHQWVRQWHQWV 1H[W&KDSWHU1H[W&KDSWHU